In the buildout phase underway, tech giants are pouring record amounts of capital into AI. Chips are the single largest cost and Nvidia is one of the big winners of that demand. Some of the most powerful chips can cost $40 billion per gigawatt versus $10-20 billion for traditional chips, according to Thunder Said Energy. Advanced chips are one reason AI data centers – the backbone of the buildout – are costlier than traditional ones. See the chart. Spending on traditional and AI data centers combined could top U.S. $700 billion annually by 2030, industry estimates show. All this spending could add to inflation, including via higher near-term energy costs given AI’s huge power needs. Eventually, AI could boost energy efficiency, offsetting some of the initial spike in energy demand. Yet those savings can come only after mass AI adoption, a process that will take time.
We think investment in AI could rival the amount seen in the industrial revolution – especially once including the spending on energy infrastructure as part of both the data center buildout and the low-carbon transition. Investment of this magnitude demands significant financing, creating a key funding role for capital markets and private markets. Yet private markets are complex and not suitable for all investors. We see big cloud providers and chip producers as the buildout’s main beneficiaries – particularly mega cap tech companies, whose unmatched resources and tech expertise give them a competitive advantage.
AI's big questions
Questions around AI overinvestment are valid. Yet we think this should be assessed in aggregate, given AI's potential to unlock new revenue streams across the whole economy. Mega cap tech does not look overextended for now. Comparisons to the dot-com era fall short, according to BlackRock’s Systematic Active Equity team: Analysis of hundreds of metrics on valuations, earnings and other features reveals few similarities between now and then. Beyond tech, other likely beneficiaries of the buildout include companies in the utilities, energy, industrials, materials and real estate sectors providing key inputs.
What comes after the buildout raises more big questions. Part of AI’s promise hinges on its ability to drive a productivity boom. Near term, we expect moderate productivity gains as AI reshapes specific tasks. Longer term, AI could accelerate the process of generating new ideas and discoveries, with far-reaching implications for innovation and growth. Much depends on how rapidly AI is adopted across industries. Broad adoption could alter the makeup of the economy by shifting labor and resources, creating new jobs and industries. Sectors like finance and IT could benefit as early adopters. If adoption happens too quickly, it could drive inflation as demand grows faster than resources can be reallocated and workers reskilled. Yet it is difficult now to imagine all of the future AI use cases. Navigating this uncertainty calls for an active investment approach, in our view. Private markets can provide an opportunity to invest in potential winners before they are publicly listed.
Our bottom line
Investment opportunities in the AI buildout expand beyond tech into sectors providing key energy, infrastructure and data center inputs. Uncertainty beyond the buildout calls for an active approach to identify future beneficiaries.
